1 Introduction

Locust plagues, floods and droughts are the three major natural disasters in Chinese history. Records from the last 2000 years described locust plagues in many scenes, such as those detailing locusts consuming all available food and resulting cannibalism. Since the 1980s, studies of natural disasters and disaster prevention activities have increased; much more attention is being paid to historical disasters and how to control future locust plagues. While studies of natural disasters can be approached from multiple perspectives, many studies have been carried out by researchers in the fields of history, geography and ecology. Geographers tend to focus on the spatial and temporal characteristics of disaster events and, to a lesser degree, case studies (event reconstruction). However, case studies are particularly lacking in prior investigations of locust plagues. In the existing studies, methods from the fields of history, sociology and natural science were combined with a geographic study of historical records to consider the temporal–spatial evolution of a major locust plague event. The related topics consider severe historical drought events (Zhang 2004a), historical documents about locust plagues (Lu 1986; Zhang 2008) and the causes and characteristics of locust plagues (Zhang 1990), the evolution of locust plague events and their relationship with society (Li 2014) as well as the effect of volcanic eruptions (Fei and Zhou 2016). Many geographers have reconstructed severe drought and flood events in different historical periods in China (1743, 1755, 1748–1787, 1876–1878 in China) based on the work of Zhang (2000, 2012), Zhang and Liang (2010) and Zhang and Demaree (2004). The extreme drought events that occurred in North China in 1876–1878 were reconstructed (Hao et al. 2010). Zhao (2005), Ma and Deng (2008), Li (2008) and Li et al. (2004) analyzed and comprehensively summarized the history of locust plagues in China; these works provided a very valuable reference for the current study.

In prior studies of the relationship between locust plagues and climate, opinions from different researchers have been mixed. Some think a warm period is a necessary condition for locust plagues to occur (Shi et al. 2014), while others feel that a cold period is required (Zhang et al. 2007). Additionally, some researchers think that rising temperatures or increasing precipitation will reduce future locust plague frequency (Ido and Guo 2013). The correlation between the occurrence of locusts and the environment has been examined from the perspective of insect ecology (Zhang 2013). However, other studies that have only predicted the possible correlation between locust plagues and global warming suggest that global warming will increase the frequency of locust plague events (Yu and Shen 2010). The reasons for this large discrepancy in opinions likely include the following: Different spatial scales led to results with varying degrees of precision. Different data sources resulted in fundamental differences in outcomes (e.g., using cited index series directly vs. reconstructing index series). Additionally, different methods have been used for analysis including space–time comparative and multiple factor analysis.

In this paper, we focus on serious locust plague events that occurred in 1850–1859 (during the reign of Emperor Xianfeng). Through analyzing the records of locust plague events and temperature–precipitation data of this period, we integrated analysis methods from the fields of history, disaster science and geography to study the spatial dynamics and temporal patterns and climate-related factors. Then, by reconstructing the spatial–temporal progression of all related events, we discuss the social impacts and climate background of the locust plague events.

2 Data and methods

2.1 Data sources

All data on locust plagues, droughts and floods during the study period were taken from “A Compendium of Chinese Meteorological Records of the Last 3000 Years” (Zhang 2004b) and combined with related local records. Reconstructed historical climate data are from the Word Data Center and the National Climate Center. Droughts, floods, locust plagues, pestilences and rice prices were summed together, and all the descriptive information including the exact times, locations and grades of disasters was recorded in detail. Then, these data were quantitatively pretreated based on previously designed semantic measurement rules. Reconstructed historical climate data included precipitation, the sunspot index, sea surface temperatures (SST), the El Niño 3 Index and global temperature series data for 1855–1859.

2.2 Methods

2.2.1 Determining the appropriate research period

According to the frequency of locust plagues during 1851–1861, we found that there was a complete locust plague duration during 1855–1859 (Fig. 1).

Fig. 1
figure 1

Locust plagues that occurred in China during 1851–1861. The actual numbers of plagues each year were 16, 11, 31, 41, 34, 201, 260, 124, 25, 16, and 10, respectively. The trend of locust plague frequency shows that one life cycle occurred in 1855–1859. The gray shaded portion indicates the major peak in locust plague events (including the beginning year, peak year and ending year.)

Full size image

According to data regarding droughts, floods, and locust for the plague cycle period, there were 644 locust outbreaks, 317 floods, 476 droughts, 203 locust plagues with droughts at the same time and 39 locust plagues with floods at the same time during 1855–1859 (Table 1).

Table 1 Total frequency of disasters during the period 1855–1859
Full size table

2.2.2 Determining the grades of disaster

Prior studies have set grades for disasters according to descriptions in historical documents (Xiao et al. 1964; Zhong and Zhao 1994). In the “Atlas of the Last 500 Years of Drought and Flood Distribution in China,” droughts and floods are divided into five grades according to historical records, though some researchers have divided them into more detailed categories (Yang and Deng 1994; Zhao et al. 2011). To determine the grades in this study, we consulted the findings of relevant prior studies (Li 2014), reading and judging comprehensively. We also compared quantitative data with qualitative data and considered the timing and duration of the locust outbreak event (e.g., one or more generations of locusts; 1 or more years), the range of the locust outbreak (local or regional) and the destruction to agriculture or impact to human beings (e.g., reduced agricultural production, famine, refugee disaster relief). With all these considerations in mind, we divided disasters into four grades; grades I–IV indicate minor, moderate, serious and severe disasters, respectively. Details of the disaster grades are given in Tables 2, 3 and 4.

Table 2 Grades of severity of historical droughts
Full size table
Table 3 Grades of severity of historical floods
Full size table
Table 4 Grades of severity of historical locust outbreaks
Full size table

Information from “A Compendium of Chinese Meteorological Records of the Last 3000 Years” and the “Atlas of the Last 500 Years of Drought and Flood Distribution in China” including locust outbreak locations and the number of counties involved as well as the grades of locust and drought severity were taken as quantitative data and divided into the above disaster severity grades.

3 Spatiotemporal progression of the locust plague

3.1 Regional features of progression

ArcGIS was used to carry out spatial analysis of locust plagues in China. The results are displayed in a map showing the precise distribution, different grades of droughts, floods and locust plagues according to county. The advantage of this method is that we can compare different disaster locations and then confirm any spatial correlations between them (Fig. 2).

Fig. 2
figure 2

Distribution of locust plagues in China from 1855 to 1859. Note: the center of the orange dots indicates the location of the locust plagues, and the size of the dots indicates the different disaster severity grades

Full size image

The distribution of locust plagues during 1855–1859 showed the following trends.

3.1.1 Distribution and migration

In 1855, locust outbreaks were mainly clustered in North China, the Huaihe River basin, South China and part of the Tibet Region. Among the outbreaks in these regions, locust plagues were more serious in North China, South China and the Yellow River and Huaihe River Basin; in contrast, locust plagues in the Tibet Region occurred in few areas and were of a minor grade. In 1856, locust outbreaks were clearly clustered in North China, the Yellow River and Huaihe River basin, while locust plagues occurring in Tibet were more diffusely distributed. In 1857, locust plagues spread to south compared with the distribution in 1856. Plagues in North China spread to Central and East China, while plagues in Tibet remained in the same locations as in the previous. In 1858, locust plagues declined throughout China, located in North China, Central China and part of East China, respectively. In 1859, locust plagues began to disappear, with only scattered distribution in North and East China.

3.1.2 Temporal evolution

From 1855 to 1859, the frequency of locust plagues followed a complete cycle: Outbreaks occurred in localized areas in 1855 and then reached a peak in 1856, followed by clustered outbreaks in North and East China in 1857. In 1858, a trend of decline began, with the same distribution as that in the peak years but a notable decrease in outbreak quantity and considerably less damage. This period began the decline of the locust outbreak, which finally ended in 1859.

3.1.3 Differences between North and South China

From the distribution patterns of all the regions in China, we identified three distinct areas (from north to south): North China, Central China and Southwest China. Locust plagues occurred more densely in the north than in the south. In the northern region, outbreaks were clustered around the Yellow River plain, Huaihe River plain and the surrounding areas. In the south, locust plagues were mainly clustered in the middle and lower reaches of the Yangtze River. During the peak years, locust plagues occurred frequently in Tibet and caused serious damage.

3.1.4 Differences between East and West China

There are no records regarding locust plagues in Northeast or Northwest China. The distribution pattern of this is similar to the population distribution pattern of Chinese (the northwest region and southeast region separated by Hu Huanyong population dividing line). This distribution illustrates that locust plagues were correlated with local agriculture and economic activities. Furthermore, this spatial distribution mirrors the division between the monsoon region and the non-monsoon region, with the main outbreak areas clustered in the monsoon region. These findings suggest that the factors that cause locust plagues are closely related to climate, not only in terms of precipitation but also related to the micro-conditions of the local area. That is, areas in the arid region have few plants to provide a food source for locusts, which was probably the main reason that few locust plagues occurred in the northwest arid zone. In contrast, in the flat southeast region, there are abundant water sources, and more highly developed agriculture, which provided more favorable conditions for locust plagues. Then, once locust plagues had occurred, dry climate will promote locust plague spread to other places.

3.2 Hot spot analysis (Getis-Ord Gi*)

In this section, we integrated the quantity and the grade of locust plagues from 1855 to 1859, taking plague grade as the weight to determine the spatial cluster patterns that occurred during this period. The results showing the distribution of hot spots and cold spots are depicted in Fig. 3.

Fig. 3
figure 3

Distribution of locust plague hot–cold spots during 1855–1859

Full size image

Figure 3 shows that the Huaihe River basin, the middle and lower reaches of the Yangtze River and the Yellow River were all hot spots. In addition, east Shaanxi Province exhibited clusters of outbreaks as well. The southwest, northwest, northeast and southeast coastal areas were all cold spot areas as the result of a smaller frequency and lower grade of outbreaks.

3.3 Cluster pattern analysis (Global Moran’s I)

Cluster pattern analysis is based on spatial autocorrelation and focused on the potential correlation between results and factors, as the first law of geography said: Anything is related to other thing else, but something close is more relevant than something in the distance. In this study, we carried out spatial autocorrelation analysis based on the locust plague grades and their locations (Fig. 4). A high–high (HH) pattern indicates a high amount of locust plague activity occurred in a region and was significantly correlated with the amounts in surrounding areas. A high–low (HL) pattern indicates that a high amount of locust plague activity occurred in the target region but the correlation with surrounding areas was not significant. A low–high (LH) pattern indicates that a low amount of locust plague activity occurred in the target region but the correlation with surrounding areas was significant. Finally, a low–low (LL) pattern indicates that a low amount of locust plague activity occurred in the target region and was not significantly correlated with the amounts in surrounding areas.

  • HH areas were mainly clustered in North and Central China in this period; this result illustrates that locust plagues caused heavy damaged and greatly impacted surrounding areas in this region.

  • The results for peak years (1856–1858) only included HH and LH patterns; this result confirmed that locust plagues trend to occur together and not independently.

  • During the first and last years of the duration (1855 and 1859, respectively), only the HH cluster pattern was present even though there was a small frequency of outbreaks, this further explained the theory that plague outbreaks only occur in tandem.

Fig. 4
figure 4

Cluster distribution of locust plagues from 1855 to 1859. NS, not significant (the cluster distribution in the region was not statistically significant); HH, high–high pattern; HL, high–low pattern; LH, low–high pattern; LL, low–low pattern. See the text for further details

Full size image

3.4 The spread and migration of plague outbreak origins

Using the grades and location information for locust plagues, we calculate the range and direction of migration in the regions studied by ellipses representing the standard deviations. Then, we calculated the mean centers of plague outbreaks from 1855 to 1859 (Fig. 5).

Fig. 5
figure 5

Characteristics of locust plague spread and mean-center migration from 1855 to 1859. Note: mean center stands for the average centers in the region studied; the yellow ellipses show the main range; and the major axis indicates the density distribution in this direction and vice visa

Full size image

The formula for the mean centers:

$$ \bar{X} = \frac{{\sum\nolimits_{i = 1}^{n} { \, x_{i} } }}{n},\quad \bar{Y} = \frac{{\sum\nolimits_{i = 1}^{n} { \, y_{i} } }}{n}. $$

In these equations, Xi and Yi stand for the latitude and longitude of the locust plague location and n stands for the quantity of locust outbreaks that occurred.

Based on the results regarding the characteristics of plague spread and mean-center migration, we determined the following:

  1. 1.

    By comparing the different years, we found that locust plagues were clustered at the beginning of the plague and then spread out from the original core locations. This trend was characterized as a “radiating after clustering” pattern with major areas of plague outbreaks in North and Central China with a mean center in Henan Province.

  2. 2.

    The locust subspecies recorded were different; the Oriental Migratory Locust was mainly clustered in East China, while the Tibetan Migratory Locust was mainly found in the Tibet Plateau. The analysis results suggest that there was no correlation between the patterns of outbreak in these two subspecies as the result of the completely different geographic and climatic backgrounds in which they are found.

3.5 Comparison analysis (droughts–floods–locust plagues)

There were three kinds of locust data: locust plagues that occurred independently, locust plagues that occurred with droughts and locust plagues that occurred with floods. For the comparison analysis, these data were combined in the same map to show their spatial relationships (Fig. 6).

  1. 1.

    During the peak years, the correlation between droughts and locust plagues was much closer than the correlation between floods and locust plagues.

  2. 2.

    Locust plague activity was mainly clustered in the heavy drought area in North China with lesser clusters in the Yangtze River basin, and near lakes and their adjacent areas in East China. This suggests that locust plague occurrence was closely associated with the drainage pattern of bodies of water.

  3. 3.

    In 1859, the locust plague frequency decreased but remained constant in drought areas.

Fig. 6
figure 6

Distribution of locust outbreaks, droughts and floods from 1855 to 1859

Full size image

4 Social impact of locust plagues

4.1 Correlation analysis over a long temporal scale

We obtained numerical locust data for 1650–1910 from historical documents and then compared the number of counties where locust plagues occurred with the number of wars per year and the rice price index (Fig. 7).

  1. 1.

    The reconstructed rice price index reflects fluctuations over this period; during 1855–1859, the price of rice increased substantially. According to the records in “A Compendium of Chinese Meteorological Records of the Last 3000 Years,” the reasons for this increase in rice prices were as follows: A long drought damaged agricultural production and famine occurred over a large area, causing rice prices to increase sharply. The drought also caused locust plagues, and locusts destroyed not only agricultural products but other food sources as well.

  2. 2.

    The frequency of wars and rice prices both decreased between the 1650s and 1690s, after which they remained somewhat stable; during 1710–1830, the number of locust outbreaks and wars stayed relatively low. However, after 1830, the number of locust outbreaks increasing suddenly and sharply, with a corresponding increase in the frequency of wars. Further analysis showed that war frequency reached a peak during 1855–1859.

Fig. 7
figure 7

Comparing locust outbreaks with wars and rice prices from 1650 to 1910

Full size image

4.2 Narrow temporal analysis

After the long-term temporal analysis, we focused on a narrower temporal analysis by reducing the time period. In the years 1855–1859, there were 7, 4, 4, 8 and 2 wars, respectively. This means that there was an average of five wars per year, which is more than three times the average for 1650–1900 (1.386 wars per year).

Figure 8 shows that locust plague frequency was positively correlated with war frequency. More specifically, wars tended to occur 1–2 years after locust plagues occurred. The result illustrates a clear hysteresis transfer effect. Furthermore, rice prices increased sharply 1/2–1 years after locust plagues occurred; again, this shows a notable hysteresis transfer effect. These phenomena suggest that locust plague directly increased the price of rice, leading to an increase in the number of refugees (Fei and Zhou 2016), which would then cause social unrest.

Fig. 8
figure 8

Comparing locust outbreaks with war and rice prices from 1840 to 1875

Full size image

5 Climate background

5.1 Regional response

Locust plague was correlated with precipitation and temperature. The results showing a strong, negative effect of precipitation and temperature affected locusts in an indirect way (Tian et al. 2011) and global warming increases the frequencies of several meteorological, agricultural and biological disasters through temperature-driven increases in insect outbreaks (Shi et al. 2014; Stige 2007).

5.1.1 Locust outbreaks and precipitation

The precipitation series in Fig. 9 was constructed using data for Huashan pine tree rings (Zhang 2004b), and the thick horizontal lines are the 10-year averages. In Fig. 9, we can clearly see the negative correlation between locust outbreaks and precipitation.

Fig. 9
figure 9

Historical locust outbreaks and precipitation from 1600 to 1900

Full size image

5.1.2 Locust outbreaks, precipitation and the Palmer drought severity index from 1851 to 1861

The Palmer drought severity index (PDSI) uses data obtained from tree rings in Henan and Shandong Province where the mean center for locust outbreaks was mainly located. The results suggest that the trends in locust outbreak frequency were opposite to the PDSI trend line. This finding reflects the fact that locust plagues are caused not only by temperature conditions but also by precipitation trends, and the negative correlation between locust plague frequency and precipitation was significant (Figs. 10, 11).

Fig. 10
figure 10

Comparison between locust plague frequency and the Palmer drought severity index from 1851 to 1861

Full size image
Fig. 11
figure 11

Comparison between locust plague frequency and precipitation from 1851 to 1861

Full size image

5.1.3 Correlation between locust outbreaks and droughts

Table 5 shows the correlation between locust outbreaks, droughts and floods. There were 208 groups of data representing locust plagues with concurrent droughts and 21 groups representing locust plagues with floods in this analysis. The results show that the correlation between locust plagues and droughts was .734, with a significant coefficient of .039 (.01 < P < .05) and a confidence interval of > 95%. This illustrates significant positive correlation existing between locust plagues and drought. The results for locust plagues with floods show that there was no significant correlation. Therefore, the main driver factor of locust plagues was drought according to the current analysis.

Table 5 Nonparametric correlation of locust plagues with drought and floods
Full size table

5.2 Global response

We took reconstructed monthly sea surface temperature (SST) from 1726 to 1996, the number of sunspots from 1700 to 2005 and the El Niño 3 index series (reconstructed based on Southern Oscillation data from 1408 to 1978) and compared these data with drought, flood and locust outbreak information from historical documents to determine the climate background of China from 1855 to 1859 (Fig. 12).

  1. 1.

    From the SST data series, surface sea temperatures exhibited a declining trend during 1855–1859 (Wang et al. 1999). This would have caused severe droughts, high atmospheric pressure, less rain and a drier climate according to the results of relevant meteorological studies (Li 2012). Thus, China was experiencing a widespread dry period during this period.

  2. 2.

    El Niño can influence the summer precipitation and temperature in China because of subtropical high pressure from the western Pacific; this would cause droughts in the north and flooding in the south (Yuan et al. 2012; Zhang and Mao 2004). Reconstructed El Niño 3 Index data show an increasing trend with fluctuation in this period; this illustrates that the temperature in North China increased as the result of El Niño 3, which would have led to severe droughts during 1855–1859.

  3. 3.

    The sunspot data series showed much more frequent sunspots and fluctuations from 1855 to 1859; this may help explain the unusual climate characteristics and frequent drought during this period.

Fig. 12
figure 12

Locust plague compared with global climate data from 1840 to 1875

Full size image

6 Conclusions

  1. 1.

    On a temporal scale, the locust plague event showed a clear 5-year life cycle, with a large fluctuation in the numbers of outbreaks per county and an overall peak in outbreaks in 1857. On a spatial scale, locust plagues were distributed more in the east than in the west and more in the north than in the south. Hot spots were located in the north, center and east of China and in areas of Shaanxi Province, while cold spots were located in the northeast, southwest, parts of the northeast and in the southeast coastal areas of China.

  2. 2.

    Locust plagues were localized at the beginning of the study period and then spread out from the original locations, presenting a “radiating after clustering” pattern. Major areas of plague were found in North and Central China and with additional cluster centers in Henan Province.

  3. 3.

    There was a close relationship between locust plagues and drought events. In addition, locust plagues were associated the drainage pattern of major bodies of water. Locust plagues in northern China were concentrated in the Huai–Hai–Yellow River plain, while locust plagues in eastern China were concentrated in the mid-lower area of the Yangtze River basin and its adjacent lakes.

  4. 4.

    A hysteresis transfer effect was observed between locust plagues and social response: Rice prices increased after locust plagues occurred with a delay of about half to 1 year and there was a clear delay of about 1–2 years between locust plague occurrence and wars.

  5. 5.

    Teleconnection occurred between climate factors that impacted the relationship between droughts and locust outbreaks, namely elevated El Niño, frequent sunspot activity and a decline in sea surface temperatures.